Sometimes the informant would tell the herald what he remembered, but memories were faulty. Sometimes he might have a pedigree written out on a fancy scroll, but the process of producing those things wasn't too scrupulous. Available data would be used, but could be misinterpreted. Then blemishes would be removed, gaps filled, extensions grafted on.
Sometimes the herald went prepared with a pedigree drafted from previous information, intending only to update it. Those trees contained errors as well.
At the end of all that, there might be two or three copies of the report, not identical. Further copies would be made. Copies went into circulation, and people made additions and alterations. The copy at the College - if there was one - wasn't immune. The heralds would doctor their records to supply ancestries for the newly-ennobled nouveau-riche.
Sometimes the editors of the books complicate things by making their own combinations from different sources. They often add extra pedigrees that didn't come from visitations at all.
So as with all "primary" sources you have to use them judiciously, bearing in mind the ways that errors creep in, and the fact that people don't always want the truth recorded in the first place.